Rankings / xAI
Grok 4
released 2025-07-10
BenchAtlas Index
as of 2026-07-14
base configuration
rank #76
10 families · 5 categories · high
Benchmark evidence
13 results
agentic coding
| τ²-Bench | 74.8 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | 81.9 % | independent | Artificial Analysis | |
| SciCode | 45.7 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | 33.3 points | independent | Artificial Analysis | |
| AA Math Index | 92.7 points | independent | Artificial Analysis |
knowledge science
| GPQA Diamond implementation=artificial-analysis | 87.7 % | independent | Artificial Analysis | |
| Humanity's Last Exam implementation=artificial-analysis | 23.9 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=artificial-analysis | 86.6 % | independent | Artificial Analysis |
long context instruction
| AA-LCR | 68.0 % | independent | Artificial Analysis | |
| IFBench | 53.7 % | independent | Artificial Analysis |
reasoning math
| AIME implementation=artificial-analysis | 94.3 % | independent | Artificial Analysis | |
| AIME year=2025 · implementation=artificial-analysis | 92.7 % | independent | Artificial Analysis | |
| MATH-500 implementation=artificial-analysis | 99.0 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model terminus-2 + Grok 4 | Terminal-Bench 1.0 | 39.0 % | community | Terminal-Bench Leaderboard |
| agent + model Artificial Analysis harness + Grok 4 | Terminal-Bench Hard | 37.9 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.